Data-Driven Remaining Useful Life Prediction of QFN Packages on Board Level with On-Chip Stress Sensors

Abstract

Miniaturization of components and higher operating loads lead to reduced lifetimes. Prognostics and Health Management (PHM) enables predictive maintenance of components whose lifetime is shorter than that of the system they are part of. The key to PHM lies in sensor data that correlates with component degradation. In this study, run-to-failure data sets have been generated using in-situ measurements of on-chip stress sensors. Physical failure analysis has provided the link between the data and remaining useful life.